ChatGPT Work logo
A

ChatGPT Work

A Tier · 8.5/10

**An Enterprise token rate card (help center, updated 2026-10-04) now prices every Work and Codex model per 1M tokens for eligible Enterprise agreements -- GPT-6 Astra $10/$50, GPT-6.1 Sol $2/$10, Daybreak Red $12.50/$75, Astra for Law $12.50/$62.50, Astra Ultrafast at 6x -- and per-token rates replace per-message billing in Chat.** EU Work output will also carry OpenAI's invisible textGrain watermark within weeks (10/05). OpenAI's long-running work agent (launched 2026-07-09) -- gathers context across your apps via plugins, stays on a project for hours, and returns finished docs/sheets/slides/web apps. **DevDay 2026 (2026-09-29) rebuilt the surface around it: dots (always-on GPT-6 Astra agents with their own cloud computer) for Pro and Business Premium with an off-by-default Enterprise beta, ChatGPT Space (which replaces Library) and Pages for Pro, Business and Enterprise, a Meetings plugin, teams with shared tasks and @ChatGPT in Slack and Microsoft Teams for Business and Enterprise, GPT-6.1 Sol in Work at near-Astra quality, and Astra Ultrafast in Work for the new $500 Pro 500 plan and Enterprise.** Earlier: GPT-6 Astra (9/03; enterprise access off by default), GPT-6 Sol and Luna (9/22), Voice inside Work (9/23). Vertical editions so far: Financial Services (9/10) and Astra for Law with 26 legal plugins (9/17).

Last updated: 2026-10-05Free tier availablePowered by GPT-6.1 Sol (added 2026-09-29, rolling out from Pro to Plus/Business/Enterprise/Edu), GPT-6 Sol / GPT-6 Luna (2026-09-22; Luna for Free/Go on desktop), GPT-6 Astra (enterprise, off by default; Astra Ultrafast on Pro 500 and Enterprise), GPT-5.6 Sol/Terra/Luna (user-selectable on paid plans); dots run on GPT-6 Astra

Score Breakdown

8.0
Ease of Use
8.5
Output Quality
8.5
Value
9.0
Features
Visit ChatGPT Work

The Good and the Bad

What we like

  • +Genuinely long-horizon: it breaks a goal into steps and works independently for hours, checking in for approvals -- this is the Codex agent loop generalized beyond code
  • +The unified plugins directory (Google Drive, SharePoint, Slack, Teams, Gmail, Outlook, Salesforce, Adobe, Zoom, GitHub, Canva, Dropbox...) plus @-mentioning a specific app gives it real context from where your work already lives
  • +Scheduled Tasks turn it into standing automation -- monitor Slack/email/dashboards on a schedule and keep docs, decks, and reports current without prompting
  • +Sites (public beta) publishes interactive dashboards, trackers, and web apps to a shareable URL -- and ChatGPT can keep them updated as the underlying data changes
  • +Desktop app goes further than web: built-in browser, local file + app access, and Computer Use that clicks/types across your desktop in the background
  • +Aggressively distributed: the desktop app with Work + Codex is free on every plan from day one

What could be better

  • −Web/mobile rollout is staged -- Plus and Business users wait days behind Pro/Enterprise/Edu at launch
  • −Usage-metered like Codex, so a long multi-hour run can eat a meaningful chunk of a $20 Plus plan's included usage -- the experience is tier-dependent
  • −An agent that reads your CRM, email, and files and then acts on them is a large trust surface -- OpenAI's own mitigation (auto-review of important actions) blocked 100% of red-team exfiltration attempts, but prompt-injection risk on connected-tool agents is structural, not solved
  • −Ships with a casualty: OpenAI is sunsetting the standalone Atlas browser in favor of the ChatGPT desktop app's built-in browser + a Chrome-sidebar extension
  • −Brand-new product surface -- expect the usual v1 roughness on long runs, plugin auth, and hand-offs between web, mobile, and desktop

Pricing

Included in ChatGPT plans (no separate price)

$0 extra
  • ✓Web + mobile: rolling out 2026-07-09 to Pro, Enterprise, and Edu first; Plus + Business 'over the next few days'
  • ✓Desktop app (Mac + Windows): Chat, Work, and Codex available on EVERY plan including Free, globally, day one
  • ✓Usage is metered like Codex -- complex tasks draw more of your plan's included usage (same structure as Codex per OpenAI's help docs)
  • ✓Free/Go users get GPT-5.6 Terra in Work (and GPT-6 Luna in the desktop app from 2026-09-22); Plus and up can pick GPT-6 Sol / GPT-6 Luna (2026-09-22) or GPT-5.6 Sol/Terra/Luna + effort level; GPT-6 Astra is enterprise-gated and off by default; `ultra` (4 parallel agents) is Pro/Enterprise in Work
  • ✓Voice inside Work on web and mobile from 2026-09-23 -- speak to create docs, decks and sheets or use connected apps; unfinished tasks continue in text after the call
  • ✓Enterprise/Edu admins get spend controls, group limits, and per-user overrides in the Admin Console

Known Issues

  • ENTERPRISE TOKEN RATE CARD FOR WORK AND CODEX -- EVERY MODEL PRICED PER 1M TOKENS, FIRST PUBLISHED DAYBREAK RED AND ASTRA LAW RATES, ULTRAFAST AT 6X (help-center article updated ~2026-10-04, vendor-primary): OpenAI's 'ChatGPT Rate Card (Enterprise token-based pricing)' covers 'eligible token-based Enterprise agreements' (credit-based workspaces keep the credit rate card until their billing model changes) and says plainly that token rates replace the old per-message rates in Chat so 'usage pricing is standardized across all product surfaces'. **ChatGPT Work and Codex models, Standard mode, per 1M input / cached input / output:** GPT-6 Astra $10 / $1 / $50; GPT-6.1 Sol $2 / $0.10 / $10; GPT-6 Sol $2 / $0.20 / $10; GPT-6 Luna $0.10 / $0.01 / $0.50; GPT-5.6 Sol $4 / $0.40 / $20 (promotional, 'available at least through Nov 21, 2026'); GPT-5.6 Terra $2 / $0.20 / $12; GPT-5.6 Luna $0.20 / $0.02 / $1.20; GPT-Rosalind-Research $5 / $0.50 / $25; GPT-5.5 $5 / $0.50 / $30; **Daybreak Blue (GPT-5.6 Sol) $4 / $0.40 / $20; Daybreak Red $12.50 / $1.25 / $75** (Daybreak needs Trusted Access for Cyber approval; Red is separately provisioned); GPT-5.3-Codex and GPT-5.2 $1.75 / $0.175 / $14; **GPT-6 Astra Law $12.50 / $1.25 / $62.50** (selected US law firms via Trusted Access). The rates apply to 'local tasks, cloud tasks, automations, code review, auto review, and delegated workers'; code review runs on GPT-5.3-Codex and auto review on GPT-5.6 Luna. **Add-ons:** Voice in Chat and Desktop $0.05 per connected minute plus backend-model tokens; web search $10 per 1,000 runs plus content tokens; GPT-Image-2.0 at $8 / $2 / $30 per 1M image tokens; deep research on o4-mini-deep-research $2 / $0.50 / $8; ChatGPT for Word, Excel/Sheets and PowerPoint and Workspace Agents at the same model rates (GPT-5.6 Sol default) sharing one agentic-usage pool. **Multipliers:** long context above 272K input tokens is 2x input, 2x cached and 1.5x output (not in the Chat tab; Astra in Codex is exempt); Fast mode 2x Standard for GPT-6.1 Sol, Astra, Sol, Luna, 5.6 and 5.5; **GPT-6 Astra Ultrafast 6x Standard under Enterprise token billing, available in Work and Codex for eligible Enterprise workspaces**; against included allowances Fast draws 2.5x and Ultrafast 8x; regional processing 1.1x. **Why it matters for buyers:** this is the first complete public Enterprise price list for Work, and the Daybreak Red and Astra Law rows had never been published anywhere.Source: OpenAI Help Center: ChatGPT Rate Card (Enterprise token-based pricing) (help.openai.com/en/articles/20001415-chatgpt-rate-card-enterprise-token-based-pricing, page reads 'Updated: 20 hours ago' on 2026-10-05) -- fetched 2026-10-05 via curl with browser UA · 2026-10-04
  • DEVDAY 2026 REBUILDS THE WORK SURFACE: DOTS, SPACE REPLACES LIBRARY, PAGES, TEAMS, @CHATGPT IN SLACK AND TEAMS, GPT-6.1 SOL, AND A $500 PRO TIER FOR ULTRAFAST (2026-09-29, vendor-primary): **(1) dots** -- 'always-on agents in ChatGPT that can take on ongoing work and keep making progress between conversations. Give your dot a goal, connect the apps it needs, and define what it can do on its own. Powered by GPT-6 Astra, your dot has its own cloud computer and can bring results back for your review.' Reach it in ChatGPT, Slack or Teams, or by voice call; it can connect to 4,000+ apps via plugins and, with permission, to your laptop. **Availability: Pro and Business Premium (18+, supported markets; Pro access excludes the EEA, Switzerland and the UK at launch); Enterprise, Edu and Healthcare get a beta that is off by default until an admin enables it.** 'For the next month, dots usage won't count toward eligible Pro, Business, and Enterprise users' plan allowances' -- usage terms come later, so there is no dots price. Specialist dots (own identity for access management, IT-provisioned hardware, systems-of-record integrations, Microsoft Agent 365) are a preview. **(2) ChatGPT Space** -- 'a new home for your team to collaborate with AI', bringing Pages, files and related work together; **'Space replaces Library for accounts with access; your existing Projects remain separate'**; Pro, Business and Enterprise on desktop and web (mobile reading and sharing now, mobile editing 'coming soon'); Enterprise sharing is an opt-in preview. **(3) Pages** -- a document type 'built for human and agent collaboration': turn a conversation into a Page, start from a template or write in Space; charts, images and interactive content; collaborators comment and edit with their own ChatGPT; sharing a Page does not expose private chats or memory. Collaborative slides with PowerPoint/Google Slides export are 'in the coming weeks'. **(4) Teams and @ChatGPT** -- Business and Enterprise can group colleagues into a team to share pages, slides, plugins and spreadsheets and delegate recurring team tasks (scheduled or event-triggered); mention **@ChatGPT in a Slack or Teams channel, thread or DM** and teammates can refine results 'without needing an individual ChatGPT license'. **(5) Meetings plugin** -- notes and personalised summaries with action items saved to Space, audio deleted once notes are ready; beta in the macOS desktop app for Pro and Business, Enterprise 'coming soon'. **(6) GPT-6.1 Sol in Work** -- 'improves on GPT-6 Sol in agentic coding, computer use, and professional work', rolling out 'starting with Pro users and expanding to Plus, Business, Enterprise, and Edu'; Enterprise and Edu owners enable it under Workspace settings > Permissions & roles. Still 'not yet available in Chat'. **(7) Pro 500 and Ultrafast** -- 'a new $500/month plan with the highest included usage of the Pro plans and access to Astra Ultrafast in ChatGPT Work and Codex' (25x the Plus allowance per the recap); Ultrafast is a faster service tier for GPT-6 Astra (up to 8x in Codex) that uses the included allowance first, then credits; among Pro plans only Pro 500 has it, and Enterprise has it too. **Pro 200 reopened to new subscriptions at $200** with changed allowances. **(8) Enterprise extras:** OpenAI Private Intelligence (Zero Data Retention with Private Safety Processing now; Private Inference preview 'this fall'), OpenAI Marketplace (apply existing OpenAI commitments to 32 partner products including Figma, Adobe, Salesforce, ServiceNow, Harvey, Legora, Palo Alto Networks, CrowdStrike and Baseten; enterprise customers 'express interest'), Sites that host plugins for Business/Enterprise/Healthcare/Edu, and shareable profiles with workspace-scoped skill sharing. **(9) The automation side:** plugins can start automations from MCP events in connected apps, and Codex Cloud tasks keep running while a laptop sleeps -- the same 'work continues without you' pattern Work launched with in July, now generalised across dots, Space and team tasks.Source: OpenAI (openai.com/index/introducing-dots/ and openai.com/index/devday-2026-recap/, both publicationDateText 'September 29, 2026'; openai.com/index/introducing-gpt-6-1-sol/, same date) + help.openai.com ChatGPT release notes 'September 29, 2026' (Pro 500, Codex Cloud, Pages, Sites, GPT-6.1 Sol in Work, dots, Meetings plugin, Space, Pro 200) -- fetched 2026-10-02 via curl with browser UA · 2026-09-29
  • GPT-6 SOL AND LUNA IN WORK, AND VOICE INSIDE WORK (2026-09-22 and 2026-09-23, vendor-primary): OpenAI's GPT-6 Sol and Luna launch says both are 'available in **ChatGPT Work** and Codex starting today for all Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. These models are not yet available in Chat.' The 9/22 release note: 'GPT-6 Sol and Luna in Work and Codex ... These models are separate from the models available in Chat ... Available models and reasoning effort options depend on your plan and workspace settings.' Why it matters for Work specifically: OpenAI's positioning for Sol is 'professional work' -- on AutomationBench 1.0.6 (47 tools across sales, marketing, operations, support, finance and HR) Sol at xhigh scores 33.2% at $0.27 per task, above GPT-6 Astra at low effort (30.3%, 3.9x the cost) and Claude Opus 5 at max (26.9%, 11.1x); on Agents' Last Exam V1 Sol at max scores 56.4%. API prices behind the metering: Sol $2/$10, Luna $0.10/$0.50 per 1M (50% below the GPT-5.6 promo rates). **Voice in Work (9/23 release note):** 'Voice is also available in Work on web and mobile. Ask it to create documents, presentations, and spreadsheets, use connected apps, or work in a browser. When you end a Voice call in Work, an unfinished task can continue in text.' Requires both Voice and Work access; existing app connections, permissions and usage limits apply. Live (Voice) also gained plugins on web, iOS and Android the same day.Source: OpenAI (openai.com/index/introducing-gpt-6-sol-and-luna/, publicationDateText 'September 22, 2026') + help.openai.com ChatGPT release notes ('September 22, 2026' and 'September 23, 2026') + developers.openai.com/api/docs/pricing.md -- fetched 2026-09-28 via curl with browser UA · 2026-09-23
  • A SECOND VERTICAL EDITION IN SEVEN DAYS: ASTRA FOR LAW AND 26 LEGAL PLUGINS (2026-09-17, vendor-primary): a week after ChatGPT for Financial Services (9/10), OpenAI introduced **Astra for Law** -- GPT-6 Astra with a legal search index (US case law, statutes, regulations, court rules and administrative decisions across more than 230 million URLs, including Free Law Project's CourtListener collection) and instructions for legal analysis and writing. For the plugins directory this page tracks, **26 partner-built legal plugins launched the same day**: iManage (draft a negotiation brief in ChatGPT and save it to the matter file), Intapp (surface activities that may need a time entry), DeepJudge (bring prior deals into a comparison), Thomson Reuters HighQ (matter context, with a CoCounsel Legal connector 'forthcoming'), plus Relativity and Clio. **The vertical is built with people as much as software:** OpenAI's forward-deployed engineers adapted ChatGPT Enterprise for **Sullivan & Cromwell** (an agreement analyzer that reads a new deal against the firm's negotiating playbooks and precedents and proposes redlines), **Ropes & Gray** (deal diligence that traces findings back to the data room) and **Cooley** (GO Public, which carries changes across an IPO filing). **Access is gated:** selected firms through **Trusted Access** in ChatGPT and Codex first, API 'coming soon', **no pricing published**; eligible firms get Zero Data Retention on the API and ChatGPT Enterprise usage excluded from human review by default, with Latham & Watkins designing information permissions, ethical walls, client instructions and firm oversight. The model appears as 'GPT-6 Astra Law' in the picker (`gpt-6-astra-law` in the API). OpenAI's own Vals AI Legal Research Bench result (54.0% overall correctness vs 38.7% for GPT-6 Astra with web search alone, highest effort, 200 private-set questions) is first-party and unreplicated. **Read-across:** Financial Services and Law are now the template -- expect further verticals packaged as data + plugins + gated access rather than as plan tiers, and priced by conversation rather than on the pricing page.Source: OpenAI (openai.com/index/astra-for-law/, publicationDateText 'September 17, 2026'; no 'Originally published on' / 'Update on' blocks) -- fetched 2026-09-17 via curl with browser UA; legal plugin tab at openai.com/business/plugins/?tab=plugins-legal · 2026-09-17
  • ADMIN ANALYTICS: USAGE, TASK INSIGHTS AND CODEX OUTCOMES IN THE ADMIN CONSOLE -- AND A PROMPT-DRIVEN ADS MANAGER PLUGIN (2026-09-16, vendor-primary): OpenAI described the analytics now in the **ChatGPT Admin Console** across ChatGPT Work and Codex. **Usage** shows active users, credits and token usage, filterable by group or user. **Insights** runs a **task classifier over a sample of messages**, grouping work into use cases and tasks (e.g. software engineering into feature development and code maintenance; sales into account research and planning) with an Overview tab and a Use cases table of credits, messages and active users per task; task details break credits down by **Models, Reasoning and Speed settings**, and a **Plugin leaderboard** and **Skills** view show which tools support each task. **Outcomes** shows Codex's contribution to **merged commits and lines of code** plus code-review activity, with group, user and repository filters. The **Admin plugin and Admin API** expose the same data for trend analysis; Datadog says it already consumes OpenAI's task categories in its Agent Console. **Governance implication:** admins can now see what people use the agent for at task granularity from a sampled classifier -- worth knowing before assuming ChatGPT Work usage is opaque to your employer. **Separately, the same day's ads post adds an Ads Manager plugin to ChatGPT Work**, letting advertisers create, update and analyse ChatGPT Ads campaigns by prompt -- the first plugin that turns Work into a channel-management console rather than a document producer.Source: OpenAI (openai.com/index/how-to-connect-ai-usage-to-business-value/, publicationDateText 'September 16, 2026'; openai.com/index/reimagining-advertising-with-ai/, same date) -- both fetched 2026-09-17 via curl with browser UA · 2026-09-16
  • THE DATA AGENT -- CHATGPT WORK CONNECTS DIRECTLY TO WAREHOUSES AND BI TOOLS AND BUILDS DASHBOARDS FROM A QUESTION (2026-09-10, vendor-primary): OpenAI introduced a **Data agent in ChatGPT Work** ('Install Data agent' is a plugin-style install). **Connections, named by the vendor:** Amazon Redshift, Datadog, Google BigQuery, ClickHouse, Databricks, MongoDB, Snowflake, Redis 'and more', plus files from Google Drive and SharePoint. **Semantic context** comes from Databricks Genie Ontology, dbt, GitHub, Snowflake Horizon and existing BI dashboards, so the agent uses the organisation's own metric definitions rather than guessing. **Output:** follow-up investigation with evidence, then an **interactive dashboard** the team can edit, share and refresh, styled to uploaded brand guidelines; it can also build and interact with dashboards in **Omni, Oracle BI, Power BI, Sigma, Tableau and ThoughtSpot** rather than replacing them. It can recommend next steps, share findings via Slack or email, and execute approved actions through connected tools. **GOVERNANCE, WHICH IS THE PART THAT DECIDES ENTERPRISE ADOPTION:** admins choose which connections exist and which roles can use them, and 'queries enforce the connected account's existing permissions, including table, row, and column restrictions' -- i.e. it inherits warehouse ACLs rather than introducing its own. **OpenAI's internal-usage claim:** 'nearly all of our product team and over two-thirds of our GTM organization use data agents in ChatGPT Work.' **READ:** this is a direct move on the natural-language-BI category (ThoughtSpot, Sigma, Databricks Genie, Snowflake CoCo), positioned as a front-end over those systems -- all six BI vendors supplied supportive quotes, which tells you OpenAI negotiated this as a channel rather than a replacement, at least for now. No separate price disclosed; it ships inside ChatGPT Work.Source: OpenAI (openai.com/index/put-data-to-work/, publicationDateText 'September 10, 2026') -- fetched 2026-09-14 via curl with browser UA · 2026-09-10
  • CHATGPT FOR FINANCIAL SERVICES -- OPENAI'S FIRST VERTICAL EDITION OF CHATGPT WORK, WITH LICENSED MARKET DATA HOSTED ON OPENAI INFRASTRUCTURE (2026-09-10, vendor-primary): OpenAI introduced **ChatGPT for Financial Services**, 'a tailored ChatGPT Work experience that combines built-in financial data with GPT-6 Astra's reasoning to help teams develop research, financial models, and customized client materials.' **Design partners:** Morgan Stanley and Evercore; initial focus is **investment banking and equity research**. **Built-in data providers:** Daloopa, PitchBook, LSEG News, Crunchbase, and Fiscal.ai -- covering earnings transcripts, financial statements, fundamentals and private-company data -- with **no separate contracts or connectors** for the built-in sets, because OpenAI 'index[es] and host[s] this data on OpenAI infrastructure.' That hosting is the differentiator versus MCP connectors: it enables **granular citations** that trace a figure back to the specific table or passage. Firms with existing subscriptions can also connect their own entitlements. Access, data connections and governance are centrally managed under ChatGPT's enterprise controls; newer models arrive 'out of the box as they are released.' **Contact-sales only, no published price.** **WHY IT MATTERS:** this is the template for OpenAI verticalising ChatGPT Work by industry (the post says the partner work 'will inform post training' and 'expansion into other financial services categories'), and it puts OpenAI in direct competition with Hebbia, AlphaSense and the banks' own copilots -- with the wrinkle that Morgan Stanley, an early OpenAI customer since 2023, is now a design partner on a product OpenAI will sell to its competitors. The data-provider quotes are vendor-selected; no accuracy benchmarks are published.Source: OpenAI (openai.com/index/introducing-chatgpt-financial-services/, publicationDateText 'September 10, 2026') -- fetched 2026-09-14 via curl with browser UA · 2026-09-10
  • GPT-6 ASTRA 'FOR WORK' -- ENTERPRISE ACCESS IS OFF BY DEFAULT, NEW ADMIN CONTROLS, AND NEW ENTERPRISE PLUGINS IN CHATGPT DESKTOP (2026-09-09, vendor-primary follow-up to the 9/03 launch): six days after launch OpenAI published a work-focused Astra post. **The operational facts for admins:** '**Enterprise access is off by default at launch**' and 'enterprise administrators can enable Astra under their applicable rate card and agreement'; new controls let them **restrict access to approved websites and desktop applications, manage uploads and downloads, and control browsing history**, on top of existing confirmation policies and automated review of unsafe tool calls. **New enterprise plugins in ChatGPT Desktop:** Oracle Analytics, Power BI (Microsoft Fabric), Navan and Avalara, 'powered by the latest browser use capabilities.' **Vendor benchmark claims (first-party):** Terminal-Bench 4.0 **57.9%** vs 37.3% for GPT-5.6 Sol and 55.8% for Claude Fable 5.1, at roughly 9% and 63% lower estimated API cost per task respectively; on OpenAI's internal computer-use safety benchmark Astra 'produced unintended outcomes 89% less often than GPT-5.6 Sol and 74.7% less often than Claude Fable 5.1'. **Pricing restated: $10 per 1M input, $50 per 1M output.** Zero Data Retention available to eligible API customers on approval. **Customer claims (vendor-selected):** Datacurve 'new record on DeepSWE v1.1 at 74%', CodeRabbit ~20% more bugs caught in review, Basis +20% pass rate on 5-hour workflows, Databricks better cost-per-task than Sol. Treat the cross-vendor comparisons as OpenAI's own measurements; Anthropic's numbers for Fable 5.1 on the same suites differ by harness.Source: OpenAI (openai.com/index/gpt-6-astra-next-generation-work/, publicationDateText 'September 9, 2026') -- fetched 2026-09-14 via curl with browser UA · 2026-09-09
  • THE ADMIN PLUGIN -- WORKSPACE ADMINISTRATION BECOMES A CONVERSATION (2026-08-25, vendor-primary): OpenAI shipped the **Admin plugin for ChatGPT Work and Codex**, letting an admin '**ask a question, explore the details, make an authorized change, and confirm the result**' in one conversation instead of moving between analytics, settings and reports. **Capabilities are genuinely write-scoped, not read-only dashboards:** review activity and credit usage across ChatGPT Work and Codex and spot members or groups approaching credit limits; **add or remove members**, update groups, run onboarding/offboarding; review effective permissions, diagnose access issues, and **control feature or model access by role or group**; and **adjust usage limits and approve or deny spending requests** against current usage. It also automates recurring workflows without custom engineering -- routing pending usage requests into **Slack or Microsoft Teams** for authorized reviewers to approve in place, and auto-granting feature access when requests meet policy. **THE THING TO WATCH IS THE BLAST RADIUS: this is an LLM with authority to change permissions, remove members and approve spend.** OpenAI's stated guardrail is that actions are 'supported' and 'authorized' and that it scales *existing* controls rather than replacing them -- but no independent audit trail or rollback story is described in the announcement, and that is the question an enterprise buyer should ask first. **Strategic read: this is ChatGPT Work's plugin directory being used for platform administration, the same vertical-channel pattern as the August education plugin bundle.**Source: OpenAI (openai.com/index/introducing-admin-plugin/, publicationDateText 'August 25, 2026'; checked for 'Originally published on' / 'Update on' blocks -- none present, this is a genuine new post) -- fetched 2026-08-28 via curl with browser UA · 2026-08-25
  • CHATGPT FOR TEACHERS ADDS 55 SCHOOL SYSTEMS AND A 16-STATE PRIVACY AGREEMENT (2026-08-26, vendor-primary): OpenAI expanded ChatGPT for Teachers to **55 additional school systems across 20 states**, reaching **over 100,000 more educators and staff**. Running totals from OpenAI: **more than 100 K-12 organizations across 30 states, over 300,000 educators and staff**, up from the roughly 150,000 at the 2025 launch. The new cohort includes **1 in 5 of America's 20 largest public school districts**. **THE STRUCTURALLY IMPORTANT PART IS NOT THE HEADCOUNT, IT IS THE PROCUREMENT UNLOCK: OpenAI announced a data privacy agreement spanning 16 states, which it calls 'a first for the industry'** -- a common framework letting districts evaluate ChatGPT for Teachers against student-data-privacy requirements without negotiating bilaterally. Student privacy review is the usual thing that stalls district AI adoption for a year, and a multi-state template is a far bigger lever on adoption than any feature. **Terms to hold onto: ChatGPT for Teachers remains free for verified U.S. K-12 educators through June 2028**, and access '**continues to be available to administrators, faculty, and educators only**' -- not students. **Read against the 2026-08-18 ChatGPT for Teens launch: OpenAI is building separate, differently-governed products for the two halves of a school building rather than one education SKU.**Source: OpenAI (openai.com/index/bringing-chatgpt-for-teachers-to-more-us-school-districts/, publicationDateText 'August 26, 2026') -- fetched 2026-08-28 via curl with browser UA · 2026-08-26
  • DATED-FLIP CHECK -- THE PREMIUM-SEAT PROMO EXPIRED WITHOUT THE SEATS EVER SHIPPING (re-verified 2026-08-20, vendor-primary): the $125 Premium seat recorded below was announced 2026-08-10 with a launch promotion that OpenAI said '**ends on August 20**'. We re-fetched the announcement on 2026-08-20, the day it lapsed. **Nothing has changed on the page**: the title is still '**Premium seats are coming to ChatGPT Business**' (future tense), the only call to action is still '**Join the waitlist**', and there is still **no GA date, no early-access list and no eligibility criteria published**. **THE PRACTICAL CONSEQUENCE, WHICH IS THE WHOLE POINT OF TRACKING THIS:** the promotion was **$100 in workspace credits per Premium seat, up to 5 seats / $500, for the first 10,000 eligible Business customers** -- but you could only qualify by **joining the waitlist**, and the seats you would spend the credits on **were never generally available during the ten days the offer ran**. So the deadline that just passed was a deadline to *register interest*, not a deadline to buy. If you joined the waitlist before today, hold OpenAI to the credits when the seats do ship; if you did not, there is currently **no published way to get them**. **FOR WORK USERS SPECIFICALLY:** the five-hour usage limit and the shared Work/Codex usage pool documented on this page are the pain the Premium seat is meant to relieve, and that relief is **still unavailable at any price** -- plan capacity on the Standard seat accordingly rather than around an imminent upgradeSource: OpenAI (openai.com/index/premium-seats-chatgpt-business/, publicationDateText 'August 10, 2026', page state re-verified unchanged 2026-08-20 via curl) · 2026-08-20
  • A $125 PREMIUM BUSINESS SEAT IS COMING, AND IT TARGETS THE USAGE-POOL PAIN ON THIS PAGE (2026-08-10, vendor-primary, WAITLIST -- not GA): OpenAI announced **Premium seats for ChatGPT Business** at **$125 per user per month, or $100 per user per month billed annually**, alongside the existing **Standard seat at $25 monthly / $20 annual**. The two features that matter for Work specifically are '**5x more usage than Standard seats**' and removal of '**the five-hour usage limit**', with predictable weekly usage resets. **Why this is relevant here and not just on the pricing page:** the loudest early complaint about Work -- recorded further down this list -- is that **Work and Codex draw from the same usage pool**, so a long Work run can silently eat your Codex limits. A 5x-usage seat you can assign to individual heavy users is the first direct answer OpenAI has offered to that, and workspaces can **mix Standard and Premium seats** rather than upgrading everyone. If you still hit a limit, shared workspace credits are available to the workspace owner. **STATUS -- DO NOT QUOTE THIS AS SHIPPED:** the post is headlined 'Premium seats are **coming** to ChatGPT Business' with a **waitlist**, and OpenAI says some customers may get early access 'before Premium becomes generally available'. There is a launch promo of **$100 in workspace credits (2,500) per Premium seat, capped at $500 / 5 seats, for the first 10,000 eligible customers, ending 2026-08-20**. Full pricing detail on /tools/chatgptSource: OpenAI (openai.com/index/premium-seats-chatgpt-business/, publicationDateText 'August 10, 2026', fetched 2026-08-10 via curl -- openai.com 403s WebFetch) · 2026-08-10
  • ROLLOUT STATUS (as of 2026-07-22): OpenAI's business release notes (7/9) say Work launched on 'all paid plans except Free and Go,' with Pro/Pro Lite/Enterprise/Edu first and 'Plus and Business users will follow over the coming days' -- no vendor confirmation of full-completion has been posted, so treat Plus/Business availability as rolling out, not formally done. **The two-week Enterprise/Edu preview -- off by default -- AUTO-ENABLES for any non-opted-out workspace on July 23, 2026** (tomorrow, as of this review): admins who want Work kept off for their org must act before the 23rd. Also confirmed at launch: ChatGPT Sites public publishing is NOT available in the EEA/Switzerland/UK, and legacy group chats began retiring 7/9. Week-two incident worth knowing (single-source, TechTimes 7/12 -- flagged as such): reports that GPT-5.6 Sol deleted user files during the Work launch, a failure mode OpenAI's own System Card documented pre-release; treat cautiously pending wider confirmation. ATLAS SHUTDOWN EXECUTED: the Atlas browser announced for sunset on 7/9 **stopped working August 9, 2026** -- now confirmed against OpenAI's own release notes ('Atlas is scheduled to stop working on August 9, 2026'), not just the earlier TechRadar report, and **verified executed on 2026-08-10** (chatgpt.com/atlas 301s to /download/; the Atlas help-centre collection 404s). Browsing/agent features folded into the unified desktop app + Chrome extension as planned. Atlas bookmarks, open tabs and history did **not** transfer automatically and the export deadline has passed -- full detail on /tools/chatgptSource: OpenAI business release notes (help.openai.com/en/articles/11391654), InfoWorld (2026, ChatGPT Work / GPT-5.6 rollout + 7/23 auto-enable), TechCrunch (2026-07-09, Atlas), TechRadar (Aug 9 stop-work date), TechTimes (2026-07-12, single-source file-deletion reports) · 2026-07-22
  • THREE EDUCATION PLUGINS FOR CHATGPT WORK AND CODEX (2026-08-04, vendor-primary): OpenAI shipped its first vertical plugin bundle, aimed at the fall term. A plugin here is '**a package of apps, role-specific skills, instructions, and common workflows**' -- the point is to skip prompt engineering, not to add a model. **Three plugins: College Student, K-12 Educator, and College Educator.** **DISTRIBUTION IS THE CATCH:** they are available '**through both ChatGPT Edu and ChatGPT for Teachers district deployments**' -- i.e. institution-managed workspaces, not individual Plus/Pro accounts. ChatGPT for Teachers is free for verified US K-12 educators and districts, so the K-12 path has no licence cost, but it does require a district or domain claim. **K-12 Educator plugin:** built with K-12 teachers to create differentiated resources, design interactive visuals and surface classroom insights; it integrates with **Learning Commons**, a philanthropic organisation building public AI datasets, which lets teachers generate material aligned to local academic standards and the learning progressions underneath them, while keeping pedagogical decisions, grading and agentic actions under teacher control. **College Educator plugin:** syllabus updates, interactive teaching sites, multimedia assessments, adapting materials for diverse learners, LMS packaging, and course calendaring against connected calendars and documents. **College Student plugin:** guided tutoring, practice on difficult concepts, study guides, quizzes, flashcards and interactive visual explanations, generated from sources the student chooses; OpenAI says it 'draws on learning science to prioritize deeper understanding' rather than answer-generation. Institutional framing, verbatim: '**AI should support learning, not shortcut it.**' Context: OpenAI is the founding partner of the **National Academy for AI Instruction** with the American Federation of Teachers, a five-year effort to train 400,000 US K-12 educators (about one in ten). Read: this is the first sign that ChatGPT Work's plugin directory is becoming a vertical distribution channel rather than a generic connector listSource: OpenAI (openai.com/index/learn-teach-chatgpt-work-codex/, fetched 2026-08-06 via curl -- openai.com 403s WebFetch) · 2026-08-04
  • ROLLOUT + USAGE-POOL GOTCHA (2026-07-10, press): the **Plus/Business rollout began 7/10** (day after Pro/Enterprise/Edu), on track with the 'coming days' guidance. Biggest early complaint (analyst-reported): **ChatGPT Work and Codex draw from the SAME usage pool** -- a long Work run can silently consume your Codex weekly limits, which surprised teams using both. Enterprise admins got per-user monthly limits + workspace overage caps in response. Also note desktop/cloud Work conversations didn't sync at launchSource: Dataconomy (2026-07-10), The Tech Portal (2026-07-10), digitalapplied.com week-one recap · 2026-07-10
  • LAUNCH (2026-07-09): OpenAI launched **ChatGPT Work**, an agent that takes goals ('analyze month-end budget variance,' 'turn research into a campaign brief') and works for hours across connected apps to return finished sheets/slides/docs/web apps. Built on **Codex technology** (5M+ weekly Codex users, 1M+ using it for non-dev work) and powered by **GPT-5.6** (GA the same day). Key surfaces: **unified plugins directory** (Drive, SharePoint, Slack, Teams, Gmail, Outlook, Salesforce, Adobe, Zoom, LinkedIn, GitHub, Canva, Dropbox), **Scheduled Tasks** (recurring/event-driven runs), **Sites in public beta** (publish interactive dashboards/web apps to a URL), and a **new unified ChatGPT desktop app** (Mac/Windows, all plans incl. Free) with built-in browser and desktop Computer Use. The **Codex app merges into the new desktop app** (developers can keep Codex as the default view); the old ChatGPT desktop app is renamed **ChatGPT Classic**; the standalone **Atlas browser will be sunset** in favor of the built-in browser + an updated Chrome sidebar extension (**this executed 2026-08-09**). Rollout: web/mobile Pro/Enterprise/Edu 7/9, Plus/Business over the following days; desktop everywhere day one. Usage follows the Codex metering structure; Enterprise admins get spend controls + Compliance API visibility. Security: 'auto-review' inspects important connected-tool actions before execution -- OpenAI says it blocked 100% of adversarial data-exfiltration attempts in red teamingSource: OpenAI (openai.com/index/chatgpt-for-your-most-ambitious-work/), Bloomberg, The Decoder, VentureBeat · 2026-07-09

Best for

Knowledge workers with recurring, multi-source deliverables -- month-end close, launch checklists, account plans, competitive research -- who already live in ChatGPT and want the agent to assemble the finished artifact rather than a draft answer. Desktop users get the most complete version.

Not for

Anyone who needs deterministic, auditable automation today (this is a v1 agent with staged rollout), teams on non-OpenAI stacks, or privacy-sensitive orgs not ready to hand an agent connected-app access. For pure coding, Codex/Claude Code/Cursor remain the sharper tools.

Our Verdict

ChatGPT Work is OpenAI generalizing the Codex playbook to all knowledge work, and it's the most credible 'agent that finishes the job' launch yet from a consumer-scale vendor: real app connections through one plugins directory, hours-long autonomous runs with approval gates, scheduled automations, and publishable Sites -- distributed free on every desktop plan from day one. The bet is distribution plus GPT-5.6's template-following gains, and the honest caveats are the usual agent ones: usage metering makes heavy runs tier-dependent, connected-tool agents carry structural prompt-injection risk, and v1 roughness is guaranteed. If you're a Pro/Enterprise user it's an immediate try; Plus users should test it the week it lands for them.

Sources

  • OpenAI Help Center: ChatGPT Rate Card (Enterprise token-based pricing) -- Work/Codex per-token rates incl. Daybreak Red, Astra Law, Ultrafast 6x, feature add-ons and multipliers (fetched 2026-10-05) (accessed 2026-10-05)
  • OpenAI: Our approach to EU text provenance rules (2026-10-05) -- textGrain watermark on EU ChatGPT and Codex output across all plans (accessed 2026-10-05)
  • OpenAI: Introducing dots (2026-09-29) -- always-on GPT-6 Astra agents; Pro/Business Premium; Enterprise beta off by default (accessed 2026-10-02)
  • OpenAI: DevDay 2026 Recap (2026-09-29) -- Space, Pages, teams, @ChatGPT in Slack/Teams, Meetings plugin, Pro 500, Marketplace, Private Intelligence (accessed 2026-10-02)
  • OpenAI: Introducing GPT-6.1 Sol (2026-09-29) -- in ChatGPT Work and Codex, not Chat (accessed 2026-10-02)
  • OpenAI Help Center: ChatGPT release notes 'September 29, 2026' (Pro 500 $500/month, Space replaces Library, Pro 200 reopened) (accessed 2026-10-02)
  • OpenAI: Introducing GPT-6 Sol and Luna (2026-09-22) -- available in ChatGPT Work and Codex, not yet in Chat; AutomationBench and Agents' Last Exam claims (accessed 2026-09-28)
  • OpenAI Help Center: ChatGPT release notes -- GPT-6 Sol and Luna in Work and Codex (Sept 22, 2026); Voice in Work and plugins in Voice (Sept 23) (accessed 2026-09-28)
  • OpenAI: Introducing Astra for Law -- GPT-6 Astra Law model, 230M-URL legal search index, Trusted Access, 26 partner legal plugins (2026-09-17) (accessed 2026-09-17)
  • OpenAI: How to connect AI usage to business value -- Admin Console Usage, Insights task classifier, Codex Outcomes, Admin plugin and API (2026-09-16) (accessed 2026-09-17)
  • OpenAI: Reimagining advertising with AI -- Ads Manager plugin in ChatGPT Work (2026-09-16) (accessed 2026-09-17)
  • OpenAI: Now everyone can put data to work -- Data agent in ChatGPT Work, warehouse + BI connections, permission inheritance (2026-09-10) (accessed 2026-09-14)
  • OpenAI: Introducing ChatGPT for Financial Services -- Morgan Stanley/Evercore design partners, hosted Daloopa/PitchBook/LSEG/Crunchbase data (2026-09-10) (accessed 2026-09-14)
  • OpenAI: GPT-6 Astra: The next generation in intelligence for work -- enterprise off by default, admin controls, new plugins, Terminal-Bench 4.0 57.9% (2026-09-09) (accessed 2026-09-14)
  • OpenAI: Introducing the Admin plugin for ChatGPT Work and Codex (2026-08-25) (accessed 2026-08-28)
  • OpenAI: Bringing ChatGPT for Teachers to more U.S. school districts (2026-08-26) (accessed 2026-08-28)
  • OpenAI: Premium seats are coming to ChatGPT Business -- $125/mo, 5x usage, no five-hour limit, waitlist (2026-08-10) (accessed 2026-08-10)
  • OpenAI: ChatGPT release notes -- 'Retiring Atlas', stops working August 9, 2026 (accessed 2026-08-10)
  • OpenAI: New ways to learn and teach with ChatGPT Work and Codex (education plugins, 2026-08-04) (accessed 2026-08-06)
  • OpenAI business release notes (Work rollout terms + 7/23 Enterprise auto-enable) (accessed 2026-07-22)
  • TechCrunch: OpenAI is shutting down Atlas (2026-07-09) (accessed 2026-07-18)
  • OpenAI: ChatGPT is now a partner for your most ambitious work (2026-07-09) (accessed 2026-07-09)
  • OpenAI: GPT-5.6 GA (powers ChatGPT Work) (accessed 2026-07-09)
  • Bloomberg: OpenAI unveils ChatGPT Work agent to field tasks for hours (accessed 2026-07-09)
  • The Decoder: OpenAI pairs GPT-5.6 rollout with ChatGPT Work (accessed 2026-07-09)

Explore more ChatGPT Work rankings

Deeper leaderboards, benchmarks, task-specific tier lists, and status/pricing pages for ChatGPT Work.

Full AI Personal Agents tier list
Where ChatGPT Work ranks vs every competitor in its category
Best AI tools to run a research agent
Autonomous agents that plan, browse, synthesize, and report on a multi-step research question.
Best AI tools to browse the web
Agents that drive a real or headless browser to click, fill, and complete multi-step web tasks.
Is ChatGPT Work down?
Outage check plus rolling log of known issues
ChatGPT Work pricing
Every tier and what's included
ChatGPT Work alternatives
Comparable tools at every tier

The Tier List Tuesday

Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.

Alternatives to ChatGPT Work

OpenClaw logo

OpenClaw

Open-source personal AI agent you talk to through Signal, Telegram, Discord, or WhatsApp. WARNING: March 2026 disclosed 9 CVEs (including CVSS 9.9) with 135,000+ exposed public instances -- verify hardening before running anywhere sensitive

B
7.6/10
Free tierFrom $0
Messaging-first interface is genuinely b...Grew from 9k to 60k+ GitHub stars in day...
Updated 2026-06-18
Hermes Agent logo

Hermes Agent

Nous Research's self-improving autonomous agent -- persistent memory, auto-generated skills, five sandbox backends. v0.18.0 'Judgment Release' (2026-07-01) adds work verification with evidence contracts, Mixture-of-Agents as selectable models, /learn + /journey self-improvement commands, and parallel background subagents

A
8.4/10
Free tierFrom $0
True learning loop -- after complex task...Five sandboxing backends (local, Docker,...
Updated 2026-07-05
Manus AI logo

Manus AI

Hosted autonomous AI agent you talk to through Telegram, WhatsApp, and Slack -- the 'no DevOps' alternative to OpenClaw and Hermes. Manus Cloud Computer (2026-04-30) adds 24/7 persistent VMs so agents keep running between sessions

B
7.9/10
Free tierFrom $0
Zero setup -- sign up, connect Telegram,...Hosted compute means the agent works eve...
Updated 2026-07-07
Perplexity Computer logo

Perplexity Computer

Perplexity's general-purpose digital worker -- operates real software like you do, runs for hours or months, **and since 2026-09-29 runs on its own: Automations replace Scheduled Tasks with schedule- or event-triggered agents (Slack, Gmail, Outlook, Linear, GitHub) that remember prior runs and can flag missed deadlines, with per-action review gates and admin-scoped connector permissions.** and since 2026-09-21 is documented as served by GLM 5.2 rather than a frontier council: Perplexity disclosed that Computer sessions run on Z.ai's open-weight GLM 5.2, post-trained on real user sessions, with the trained checkpoint offered as a model option (21.2% fewer live tool-call failures). **Portable Computer** runs the whole thing on your own hardware -- NVIDIA RTX since 8/25, Windows since 9/14, AMD Ryzen AI Max since 9/24 -- on PPLX 27B or Qwen 3.8 27B with 24 GB of GPU memory, and local inference does not consume Computer credits. Included on Pro ($20) since 5/07, not only Max ($200).

A
8.4/10
From $20
Best-in-class model routing -- it uses O...Truly long-running -- workflows can run ...
Updated 2026-10-02
Wingman (Emergent) logo

Wingman (Emergent)

Emergent's messaging-first personal AI agent -- launched 2026-04-15 from the India vibe-coding startup ($70M raise, $300M valuation). Positioned as an OpenClaw alternative with safer defaults

A
8.1/10
Free tierFrom $0
Same-day fresh -- launched 2026-04-15, b...WhatsApp + Telegram first-class, same UX...
Updated 2026-04-17
Perplexity Comet logo

Perplexity Comet

Perplexity's agentic AI browser -- FREE on all platforms as of 2026-03-18 (previously $200/mo Max-only). iOS, Android, Windows, Mac. Browses the web, executes multi-step tasks, and summarizes pages in-line. Comet Plus ($5/mo) adds premium publisher content

A
8.4/10
Free tierFrom $0
Went FREE on all platforms 2026-03-18 --...Agentic browsing does real work -- books...
Updated 2026-04-17